Forward-propagation rule based on ridge regression for inverse kinematics problem

We consider solving the inverse kinematics problem by forward-propagation rule with high order term. The goal signal is derived by the Newton-like method and the correction of weights is calculated by ridge regression. Hence, the learning times would be reduced if we can realize the goal signal accurately because this signal is derived by Newton-like method. We propose adjusting the regularization parameter of ridge regression depending on the high order term. The experimental result shows decrease of the learning times without loss of the accuracy of the inverse kinematics model.